Chinese Tech Giants Shift to Domestic AI Chips Amid Nvidia Export Curbs

Chinese technology giants such as Alibaba, Tencent, and Baidu are accelerating their shift toward domestically produced AI chips, prompted by tightening U.S. export controls and a dwindling supply of Nvidia processors. The Trump administration’s recent restrictions on Nvidia’s H20 chip have intensified the urgency for China to develop self-sufficient AI infrastructure.

Key Developments:

• Transition to Domestic Chips: With existing Nvidia supplies projected to last only until early next year, Chinese tech companies are testing local alternatives like Huawei’s Ascend chips. However, transitioning from Nvidia’s CUDA to Huawei’s CANN framework presents technical challenges and could cause development delays of up to three months.
• Hybrid Strategy: Companies are adopting a hybrid approach, using remaining Nvidia chips for AI training while shifting to domestic processors for inference tasks. This strategy aims to balance performance needs with the realities of supply constraints.
• Supply and Demand: Huawei is expanding production, though current supply does not meet demand. Other chipmakers like Cambricon and Hygon are also being explored, and some tech companies are investing in building their own processors.
• Regulatory Concerns: The use of Huawei’s Ascend chips has raised concerns due to U.S. warnings that global use may incur penalties. This has led companies to remain discreet about testing and deploying these chips.
• Nvidia’s Response: Nvidia may release a compliant version of its Blackwell chip—without critical features like high-bandwidth memory—as early as July. However, it’s unclear whether this will meet the performance needs of Chinese tech firms.

This strategic pivot underscores China’s commitment to achieving technological self-reliance in the face of escalating trade tensions and export restrictions.

For more details, read the full article on the Financial Times: Chinese tech groups prepare for AI future without Nvidia.